Method and apparatus to correct digital image blur due to motion of subject or imaging device
View Patent ↗Signal processing techniques are applied to digital image data to remove the distortion caused by motion of the camera, or the movement of the subject being photographed, or defective optics, or optical distortion from other sources. When the image is captured, the effect of relative motion between the camera and the subject is that it transforms the true image into a blurred image according to a 2-dimensional transfer function. The 2-dimensional transfer function representing the motion is derived using blind estimation techniques or by using information from sensors that detect the motion. The transfer function is inverted and used to define a corrective filter. The filter is applied to the image and the blur due to the motion is removed, restoring the correct image. Another embodiment uses the transfer function to avoid blur by combining multiple consecutive images taken at a fast shutter speed.
1. A method, comprising:
a recording medium recording image data that represents an image;
a processor determining an initial state for an inverse transfer function;
the processor updating the inverse transfer function using an iterative technique, wherein the iterative technique includes, for each of a plurality of iterations:
determining a first estimate of the image data based on the inverse transfer function and a finite impulse response (FIR) filter;
determining a second estimate of the image data by applying a non-linear estimation function to the first estimate of the image data;
determining an estimation error parameter by comparing the second estimate of the image data and the image data; and
updating the inverse transfer function based on the estimation error parameter and a step-size parameter;
the processor determining a deconvolution filter based on the inverse transfer function; and
the processor filtering the image data using the deconvolution filter.
2. The method of claim 1 , wherein the iterative technique further includes, for one or more of the plurality of iterations:
determining whether to perform a next iteration based on the estimation error parameter.
3. The method of claim 1 , wherein the inverse transfer function includes one or more coefficients, and wherein the updating the inverse transfer function includes updating the one or more coefficients.
4. The method of claim 1 , wherein the second estimate of the image data is a Bayes estimate.
5. The method of claim 1 , wherein the transfer function represents motion in a portion of the image data that is smaller than the entire image data, and wherein the processor filtering the image data using the deconvolution filter is performed on the entire image data.
6. The method of claim 1 , wherein the method is performed in a digital camera.
7. An apparatus, comprising:
a recording medium configured to record image data that represents an image;
a processor configured:
to determine an initial state for an inverse transfer function;
to update the inverse transfer function using an iterative technique, wherein the iterative technique includes, for each of a plurality of iterations:
determining a first estimate of the image data based on the inverse transfer function and a finite impulse response (FIR) filter;
determining a second estimate of the image data by applying a non-linear estimation function to the first estimate of the image data;
determining an estimation error parameter by comparing the second estimate of the image data and the image data; and
updating the inverse transfer function based on the estimation error parameter and a step-size parameter;
to determine a deconvolution filter based on the inverse transfer function; and
to filter the image data using the deconvolution filter.
8. The apparatus of claim 7 , wherein the iterative technique further includes, for one or more of the plurality of iterations:
determining whether to perform a next iteration based on the estimation error parameter.
9. The apparatus of claim 7 , wherein the inverse transfer function includes one or more coefficients, and wherein the updating the inverse transfer function includes updating the one or more coefficients.
10. The apparatus of claim 7 , wherein the transfer function represents motion in a portion of the image data that is smaller than the entire image data, and wherein the processor filtering the image data using the deconvolution filter is performed on the entire image data.
11. The apparatus of claim 7 , wherein the second estimate of the image data is a Bayes estimate.
12. The apparatus of claim 7 , wherein the apparatus is a digital camera.